memlab
A framework for finding JavaScript memory leaks and analyzing heap snapshots
by facebookTypeScript
Commit activity
Last 12 weeks · 54 commits
Community health
4 of 6 standards met
75
✓README✓License✓Contributing✓Code of Conduct○Issue Template○PR Template
Recent PRs & issues
Active · 2 in progress · Last activity 19h ago
See all on GitHub →
Hi memlab devs, thanks for the great tool. I'm excited to try to integrate it into my current NodeJS app tests. I'm wondering what's current status or future road map is on supporting NodeJS memory leak detection. Currently, most test cases and demos are running in browsers detecting front-end side memory leaks. Would you plan to give more support on NodeJS?
Summary New package — a Playwright fixture that attaches memlab's leak detection to existing Playwright tests by destructuring a parameter. No scenario rewrite, no Puppeteer dependency. Motivation memlab's existing APIs (, ) drive the browser through a Puppeteer-based scenario runner. Teams already running Playwright e2e suites can't attach leak detection without rewriting their specs. This package bridges the gap by piggybacking on the user's existing page via a CDP session. Design notes Chromium only. Heap snapshots go over CDP, which Playwright only exposes for Chromium. On Firefox / WebKit the fixture becomes a no-op: the test still runs, no leak detection, a memlab-skip annotation is attached. README calls this out. Opt-in per test. Fixture setup runs only when memlab is destructured. Tests that don't use it pay no overhead 3-phase snapshot (baseline / target / final) wired to fixture lifecycle. via , no flag required. Configurable. for per-project GC tuning and custom filters. Merge semantics preserve prior fields. CI-friendly failure output. picks the hop (not the Window root), strips memlab-internal metadata from attached JSON, and heap snapshots only attach when leaks are detected (skips ~13 MB per clean test). API : Capture baseline heap snapshot / : Manual phase control : Per-test tuning : Returns : Hard assertion with triage-friendly error If neither nor is called, the fixture auto-captures missing phases and soft-asserts no leaks during teardown.
I was so excited when I read about this project ... until I realized it was for Pupeteer :( Any chance you'd consider adding Playwright support for this library?
Recent fixes
I was looking around memory leak detection and stuff and came across memlab. I looked in the docs and have a basic idea of how memlab works, but I had a question. When memlab receives the three snapshots namely baseline, target and final, it generates a superset of leaked object by the formula (T-B) ∩ F. Now when we run the find-leaks command and pass in these 3 snapshots, does memlab internally also mark each fiber node as detached by following its retainer path and finding whether it is connected to a root fiber node or not. In the chrome devtools I do not see the word Detached beside the FiberNode objects, how does memlab decide which fiber nodes are detached and which are not?
Node count in one of the heap snapshots is 16,997,604 which exceeds the maximum size of 16,777,216. For context, this started happening after upgrading Chrome to M148 which seems to include a lot more nodes in heap snapshots. Assuming this is an intentional change with Chrome, I'm wondering if it's possible to use a different data structure that can accommodate larger node counts.
Fixed a typo in the documentation, changing "teesting" to "testing" in the E2E frameworks integration guide.
Structured data for AI agents
Repository: facebook/memlab. Description: A framework for finding JavaScript memory leaks and analyzing heap snapshots Stars: 4989, Forks: 142. Primary language: TypeScript. Languages: TypeScript (54%), JavaScript (45.3%), MDX (0.4%), CSS (0.2%), HTML (0.1%). License: MIT. Homepage: https://facebook.github.io/memlab/ Topics: detector, e2e, facebook, heap, hermes, javascript, leak, memory, nodejs, perf-tools, performance, snapshot, v8. Open PRs: 2, open issues: 11. Last activity: 19h ago. Community health: 75%. Top contributors: JacksonGL, tulga1970, Drizzlecrj, mrsharpoblunto, LuciNyan, aelij, kleookku, arunsathiya, eltociear, mehmetmalli and others.